Recently, wireless mesh networks attract more and more attention, e.g. for remote control of illumination systems, building automation, monitoring applications, sensor systems and medical applications. In particular, a remote management of outdoor luminaires, so-called telemanagement, becomes increasingly important. On the one hand, this is driven by environmental concerns, since telemanagement systems enable the use of different dimming patterns, for instance as a function of time, weather conditions and season, allowing a more energy-efficient use of the outdoor lighting system. On the other hand, this is also driven by economical reasons, since the increased energy efficiency also reduces operational costs. Moreover, the system can remotely monitor power usage and detect lamp failures, which allows determining the best time for repairing luminaires or replacing lamps.
Current radio-frequency (RF) based wireless solutions use either a star network topology or a mesh network topology. In a star network, a controller has a direct communication path to every node in the network. However, this typically requires a high-power/high-sensitivity base-station-like controller to be placed at a high location (e.g. on top of a building), which makes the solution cumbersome to deploy and expensive. In a mesh network, the plurality of nodes does in general not communicate directly with the controller, but via so-called multi-hop communications. In a multi-hop communication, a data packet is transmitted from a sender node to a destination node via one or more intermediate nodes. Nodes act as routers to transmit data packets from neighboring nodes to nodes that are too far away to reach in a single hop, resulting in a network that can span larger distances. By breaking long distances in a series of shorter hops, signal strength is sustained. Consequently, routing is performed by all nodes of a mesh network, deciding to which neighboring node the data packet is to be sent. Hence, a mesh network is a very robust and stable network with high connectivity and thus high redundancy and reliability.
In the prior art, mesh network transmission techniques can be divided in two groups: flooding-based and routing-based mesh networks. In a flooding-based mesh network, all data packets are forwarded by all nodes in the network. Therefore, a node does not have to make complicated routing decisions, but just broadcasts the data packet. By these means, the technique is quite robust. However, in large networks, the data overhead due to forwarding impacts the overall data rate. Moreover, collisions of data packets are more likely to occur, further reducing the overall performance. Hence, the main problem of this solution is the scalability. Routing-based mesh networks can be further divided into proactive and reactive schemes. In proactive routing-based mesh networks, all needed network paths are stored in routing tables in each node. The routing tables are kept up to date, e.g. by sending regular beacon messages to neighboring nodes to discover efficient routing paths. Although the data transmission is very efficient in such kind of network, the scalability is still low, since in big networks, the proactive update of the routing tables consumes large parts of network resources. Moreover, the routing tables will grow with the scale of the network. In addition, the setup of the network requires time and resources in order to build up the routing tables. Reactive schemes, in contrast, avoid the permanent overhead and large routing tables by discovering routes on demand. They use flooding to discover network paths and cache active routes or nodes. When routes are only used scarcely for single data packets, flooding the data packets instead of performing a route discovery might be more efficient. If routes are kept long enough to avoid frequent routing, reactive schemes degenerate to proactive schemes. An example for a reactive routing-based mesh network is used in ZigBee. However, the main problem of this protocol scheme is still the scalability of the network.
Thus, most transmissions in wireless mesh networks are performed in a multi-hop mode. Since then every data packet or message is transmitted multiple times due to the forwarding, thereby reducing the overall network throughput, the network scalability of wireless mesh networks is very limited. Also, data packet collisions are more likely to occur, further degrading the overall performance. In particular, when node-initiated data traffic, e.g. statistics report data or other time-uncritical data, is transmitted by a plurality of nodes to a data collector node or control center choosing nearly the same moment in time, an overload problem will arise, causing data collision and loss.
Moreover in non-reliable networks, such as RF networks, a data packet can also get lost during transmission for other reasons than collisions, e.g. due to network overload or deteriorated link quality. Retransmissions can reduce the likelihood, but never can guarantee a successful transmission. The likelihood of packet losses adds up, when a packet has to travel over multiple hops. In large-scale multi-hop networks, the number of hops a data packet has to travel is comparatively large. For instance, in a large RF telemanagement system comprising thousands of nodes, 20-40 hops are likely to occur. Hence, the delivery chance of an individual data packet decreases with its hop distance, since with every hop, there is a chance that the data packet gets lost. Moreover, due to congestions and temporary errors at routing level, the likelihood of packet losses in the multi-hop scenario increases further. This makes data acknowledgment at transport or application layer necessary, if delivery guarantees are required by the application. The drawback of an acknowledge mode, however, is that data acknowledgements increase the data load in the network and the experienced delay increases significantly, especially when data packets have to be retransmitted at transport or application layer. Moreover, when a multicast or broadcast data packet is responded to (acknowledged) by many or all receiver nodes within a short period of time, a so-called acknowledgement storm may occur, causing an overload problem in the neighborhood of the sender node and thus in data collision and loss. This can be avoided in globally scheduled networks, where every node has fix individual timeslots for transmissions. But this global schedule needs global coordination and configuration, thus involving a cumbersome set-up procedure. Furthermore, sending over a global schedule and creating a global schedule for all nodes create a high data overhead and management overhead, respectively. Moreover, when the slots are used only infrequently, this decreases drastically the possible bandwidth. Consequently, this approach is also not suitable for large-scale networks.
Hence, a big disadvantage in common wireless networks is constituted on the one hand due to the tedious deployment and configuration and on the other hand by the very limited network scalability. Especially, RF telemanagement networks suffer from significant overload due to their topology and size, particularly at local level, which limits their scalability. This occurs because messages are generated and transmitted by layers of the communication stack without considering the network topology. Improving the success and reliability of transmissions is therefore crucial in large-scale networks, such as street illumination systems with a high number of luminaire nodes. This is because end-to-end retransmissions typically occurring at the higher layers in the communications stack are far more costly and delay intensive. Consequently, efficient routing protocols are required for large-scale wireless networks in order to minimize data packet collisions and subsequent data loss and to achieve the required throughput, response times and robustness.
US 2006/0187836 A1 describes a communication device that delays or discards time-insensitive traffic prior to time-critical traffic by introducing a virtual bottleneck queue.
Whetten et al. (IEEE Network, vol. 14, no. 1, pages 37-47, XP001195288) describes the Reliable Multicast Transport Protocol II (RMTP-II). A receiver node periodically informs its parent node about the packets it has or has not received by unicasting a TRACK (hierarchical ACK) packet to its parent node. Each parent node aggregates the TRACKs from its child nodes and unicasts a single aggregated TRACK to its parent (top) node. The top node aggregates the TRACKS from its child nodes and unicasts a single TRACK to the sender node.
US 2010/246600 A1 describes a system for reducing overhead in a communication system. Different frame types are aggregated into an enhanced aggregated frame in response to a determination that the different frames are suitable for transmission within a single reservation period. If the frames are time-sensitive, the transmitter, or transceiver, will not aggregate the frames as it needs to flush out the frames as quickly as possible.
US 2008/137624 A1 describes a global wireless sensor network architecture and protocol for remote supervision, asset control and operational management of ad hoc networks. An application of this technology is the remote control and supervision of lighting systems from centralized regional operations centers.